基于多目标不确定性改进的GM-PHD滤波器  

Improved GM-PHD Filter Based on Multi-target Uncertainty

作　　者：王奎武 张秦[1] 虎小龙 WANG Kuiwu;ZHANG Qin;HU Xiaolong(School of Air Defense and Missile Defense,Air Force Engineering University,Xi an 710051,Shaanxi,China;Graduate School,Air Force Engineering University,Xi an 710051,Shaanxi,China)

机构地区：[1]空军工程大学防空反导学院,陕西西安710051 [2]空军工程大学研究生院,陕西西安710051

出　　处：《兵工学报》2022年第12期3113-3121,共9页Acta Armamentarii

基　　金：陕西省自然科学基础研究计划项目(2022JQ-679)。

摘　　要：基于随机有限集的高斯混合概率假设密度(GM-PHD)滤波是处理多目标跟踪问题的一种有效方法。GM-PHD滤波器在密集杂波环境中会因估计误差过大而导致跟踪性能的下降,主要是因为没有充分考虑来自多目标量测的不确定性。为此,提出在考虑高斯分量权重的情况下,通过分量值改变协方差更新式,并通过引入标签,采用自适应阈值对高斯分量进行合并。理论分析和仿真结果表明:该方法在杂波环境下,目标最优次模式分配距离小,跟踪精度更高;目标数量的估计结果受杂波的影响更小,其估计值更接近真实的目标数量;通过具有不同杂波以及检测概率条件的跟踪场景,证明了该方法的目标数量估计精度和滤波性能明显好于传统算法。Gaussian mixture probability hypothesis density(GM-PHD)filtering based on random finite set(RFS)is an effective method for solving multi-target tracking(MTT)problem.However,the GM-PHD filter in a dense clutter environment may have a poorer tracking performance due to excessive estimation errors.This is mainly because the uncertainty from the multi-target measurement is not fully considered.Every moment,the GM-PHD filter generates a new Gaussian component and a specific measured value.When clutter density is high or the detection probability is low,the accuracy of the estimated value of the target is lower.To solve this problem,this paper proposes a change in the covariance update formula through adjusting the component value with the weight of the Gaussian component considered,and introduces a label to merge the Gaussian components with an adaptive threshold,so as to improve the accuracy of the target number estimation.The proposed algorithm is proven to have a significantly higher target number estimation accuracy and filtering performance than traditional algorithms through tests under tracking scenarios with different clutter and detection probability conditions.

关 键 词：多目标跟踪 随机有限集 高斯混合概率假设密度滤波器 高斯混合 状态估计 

分 类 号：V556[航空宇航科学与技术—人机与环境工程] TP391[自动化与计算机技术—计算机应用技术]